A Machine-learning Approach for Video Streaming Provisioning at the Network Edge

5G is the 5th generation mobile network, the planned successor to the 4G networks which provide connectivity to most current cellular devices .5G enables a new kind of network that is designed to connect virtually everyone and everything together, including machines, objects, and devices. The new mobile network provides greater bandwidth, giving higher download speeds, up to 10 gigabits per second. Owing to higher bandwidth than before, it is expected that the new networks will not just serve cellphones like before, but also serve for laptops and desktop computers. The workload of the network data center increases significantly, especially in the edge data center in which the resource is limited. This thesis aims to create an AI model for optimal resource allocation, which controls the workload to avoid overload in the edge data center. There are different types of services, e.g AR/VR (Augmented Reality and Virtual Reality), V2X (Vehicle-toEverything), Live Video Streaming, manufacturing and health. Failure in safety service will cause damage to human lives, while failure in entertainment service will affect the user experience. Live Video Streaming is the primary concern of this thesis. Live streaming is when the streamed video is sent over the Internet in real-time, without first being recorded and stored. There are a lot of steps which take place behind the scenes in a live stream and only the encoding part will be investigated. The final objective of this study seeks to create a machine-learning algorithm to control the encoding process in Live Streaming to avoid latency due to overload.